Citation

Izani, M. and Kaleel, Akhmed and Assad, Amr and Bahrin, Kamal and Abdul Razak, Aishah and Rosli, Mohd. (2025) Measuring the Performance of AI-Generated Animations: Strengths, Weaknesses, and Future Directions. In: 2025 22nd International Learning and Technology Conference (L&T), 15-16 January 2025, Jeddah, Saudi Arabia.

Text 14.pdf - Published Version Restricted to Repository staff only Download (592kB)

Abstract

Generative AI has rapidly advanced in producing visually appealing animations, but its ability to capture the subtle complexities of traditional animation techniques remains insufficiently explored. This study provides a detailed evaluation of AI-generated animations through the lens of the 12 classical principles of animation, which are essential for creating lifelike and engaging motion. Utilizing expert evaluations combined with robust statistical methods, including Chi-square tests and logistic regression, we quantitatively compare AI-generated animations to traditional hand-crafted ones. Our findings reveal that while AI-generated animations perform well in static principles such as Staging, Solid Drawing, and Appeal, they consistently fail to replicate dynamic principles like Squash and Stretch, Follow-Through, and Anticipation. This analysis represents the first quantitative benchmark of AI's performance in animation, offering critical insights into its strengths and shortcomings. We conclude with specific recommendations for improving generative AI models, particularly in motion dynamics, and outline future research directions to bridge the gap between AI-driven and traditional animation techniques.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Generative AI, principles of animation, statistical analysis, traditional animation comparison
Subjects:	N Fine Arts > N Visual arts
Divisions:	Faculty of Creative Multimedia (FCM) > Knowledge Management Centre
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	30 Apr 2025 03:45
Last Modified:	30 Apr 2025 03:45
URII:	http://shdl.mmu.edu.my/id/eprint/13726

Downloads

Edit (login required)